Sexual deception in a cannibalistic mating system? Testing the Femme Fatale hypothesis

Abstract

Animal communication theory holds that in order to be evolutionarily stable, signals must be honest on average, but significant dishonesty (i.e. deception) by a subset of the population may also evolve. A typical praying mantid mating system involves active mate searching by males, which is guided by airborne sex pheromones in most species for which mate-searching cues have been studied. The Femme Fatale hypothesis suggests that female mantids may be selected to exploit conspecific males as prey if they benefit nutritionally from cannibalism. Such a benefit exists in the false garden mantid Pseudomantis albofimbriata—females use the resources gained from male consumption to significantly increase their body condition and reproductive output. This study aimed to examine the potential for chemical deception among the subset of females most likely to benefit from cannibalism (poorly fed females). Females were placed into one of four feeding treatments (‘Very Poor’, ‘Poor’, ‘Medium’ and ‘Good’), and males were given the opportunity to choose between visually obscured females in each of the treatments. Female body condition and fecundity varied linearly with food quantity; however, female attractiveness did not. That is, Very Poor females attracted significantly more males than any of the other female treatments, even though these females were in significantly poorer condition, less fecund (in this study) and more likely to cannibalise (in a previous study). In addition, there was a positive correlation between fecundity and attractiveness if Very Poor females were removed from the analysis, suggesting an inherently honest signalling system with a subset of dishonest individuals. This is the first empirical study to provide evidence of sexual deception via chemical cues, and the first to provide support for the Femme Fatale hypothesis.

1. Introduction

The handicap principle [1] suggests that costly signals will always be honest because only the best-quality individuals will be able to produce the best-quality signals. However, more recently it has been suggested that signalling systems will rarely, if ever, be perfectly honest and that infrequent dishonest signalling can be an evolutionarily stable strategy (ESS) within a system of overall signal honesty [2–5]. Both the cost of signalling and the benefits to be gained by signalling can influence signalling strategy [2,3,5,6]—a given level of signal may indicate a high-quality individual that has less to gain from signalling but can easily bear the costs, or a poor-quality individual that has more to gain from signalling but cannot easily bear the costs [2,3,5]. While there are many examples of intraspecific deception in the context of mate competition [7–13], there are relatively few illustrations of intraspecific deception via mate choice [14,15]. However, one such example comes from the long-tailed dance fly (Rhamphomyia longicauda), where females deceive males by signalling the readiness of eggs that are actually immature, thereby obtaining a protein-rich nuptial gift in exchange for a copulation that leads to relatively low paternity for the male [15].

Sexually cannibalistic mating systems provide an opportunity to study the reliability of signals within the framework of extreme sexual conflict. In both cannibalistic spiders and mantids, females chemically signal their receptivity to mate-searching males [16–25], a mechanism by which females could potentially exploit sexually motivated males as easy prey (the Femme Fatale hypothesis [26,27]). This may be particularly relevant in species that gain nutritional benefits from cannibalism (see table 1 in [28]), and for individuals within these species that are food limited and/or in poor nutritional condition. In praying mantids, the fitness cost of this deception for males is potentially severe—food-limited females are not only less fecund [18,21,29] but also significantly more likely to cannibalize than well-fed females ([21,28,30–32]; table 1). In spite of the Femme Fatale hypothesis, honest signalling systems have been demonstrated in all cannibalistic mantid species studied thus far [18,20–22,33,34]. The overwhelming consensus is for complete differential attraction to females in good condition; that is, well-fed females are chemically attractive to males, whereas hungry females in poor condition are almost exclusively unattractive. A recent study of Pseudomantis albofimbriata [18] delved deeper into the relationship between female body condition and attractiveness, and found evidence of a more complex scenario in which egg production and pheromone production are linked. First, many poor-condition females with eggs in their ovaries were attractive to males, but poor-condition females with no eggs were never chemically attractive to males. Second, attractive females were significantly more fecund than unattractive females within both the poorly fed and well-fed treatments, but body condition per se did not have an effect on attractiveness within treatments. Third, attractive poor-condition females were still attractive in the presence of good-condition females, suggesting that all females can potentially attract males if they have a base level of fecundity. These results suggest that previous findings of complete differential attraction in mantids [18,20,21] may be owing to food-limited females being physiologically incapable of producing eggs and, therefore, pheromone [35,36].

Table 1.

Comparison of female traits between the four feeding regimes (superscript letters (A, B and C) represent Tukey's values).

treatment	Good (n = 6)	Medium (n = 6)	Poor (n = 6)	Very Poor (n = 6)
body size (mm)	15.505 ± 0.199A	15.882 ± 0.221A	15.842 ± 0.236A	15.522 ± 0.354A
body condition (mm g−1)	0.062 ± 0.005A	0.057 ± 0.002A,B	0.049 ± 0.001B	0.027 ± 0.001C
fecundity (no. eggs)	62.33 ± 7.11A	52.67 ± 2.45A	32.83 ± 2.94B	5.33 ± 1.65C
frequency of cannibalism	low—0% ([28], n = 19)a	low—0% ([28], n = 19)a	medium—42% ([28], n = 35)b	high—90% ([28], n = 19)b
frequency of copulation	very high—100% ([28], n = 19)a	very high—100% ([28], n = 19)a	high—79% ([28], n = 35)b	medium—50% ([28], n = 19)b

^aSimilar feeding regime and subsequent cannibalism frequency reported (one large cricket instead of many small crickets).

^bExact feeding regime and subsequent cannibalism frequency reported.

This study of the false garden mantid P. albofimbriata aims to overcome the aforementioned constraint by assuring that food-limited females in poor condition have begun egg production (and are therefore theoretically able to produce pheromone [18]), and subsequently determine whether these females deceptively signal to males. To identify cases of deceptive signalling, overall honesty in the population must first be determined by establishing a correlation between the signal and some characteristic of the quality of signaller. Once this correlation is ascertained, cases where the correlation is lost in ways that benefit the signaller can be identified [5]. In this study, female attractiveness (i.e. total number of males attracted) is used as a proxy for pheromone quality and/or quantity and female fecundity is used as the related female characteristic.

Three mutually exclusive outcomes are possible: (i) no pattern—attractiveness is unrelated to female condition/fecundity, e.g. all egg-bearing females are similarly attractive irrespective of egg number; (ii) total honesty—attractiveness increases with increasing food quantity and positively relates to female condition/fecundity across all treatments; (iii) partial dishonesty within overall honesty—attractiveness generally increases with increasing food quantity and positively relates to female condition/fecundity, with the exception of the poorest-quality females. In this scenario, poor-quality females fit Johnstone and Grafen's description of poor-quality individuals that have more to gain from being deceptive but cannot easily bear the costs involved [2], and therefore produce a quantity or quality of pheromone that mimics a high-quality/low-risk individual instead of a poor-quality/high-risk individual.

2. Material and methods

(a). Study species and site

Individual P. albofimbriata (N = approx. 100) were collected in January 2012 from various sites around Sydney, Australia. The majority of individuals were found in Lomandra longifolia bushes at Kuringai Bicentennial Park, West Pymble, Sydney (33°45′37.76″ S, 151°08′20.88″ E). Juvenile mantids (in their antepenultimate or penultimate instar) were collected from the study sites and maintained on a diet of two small crickets (Acheta domestica: mean cricket body mass = 0.034 ± 0.002 g, N = 50) three times a week and sprayed with water daily. Animals were housed individually within well-ventilated 425 ml transparent cups in the laboratory, at a temperature of 24–26 °C and with a diurnal period of 14 light hours per day.

(b). Measuring and sexing individuals

The pronotum length of all mantids was measured after the final moult using electronic callipers and was used as a measure of fixed adult size, whereas body mass was measured immediately preceding mate choice experiments using electronic scales. I used body mass divided by fixed size and the residuals of a regression of body mass over fixed size as indices of body condition [37]. However, both gave very similar results, so I report only fixed size over body mass throughout this paper. The sex of P. albofimbriata individuals was determined by differences in the adult abdomen and wing morphology. The total number of eggs in a female's ovaries was used as a measure of her fecundity.

(c). Feeding treatments

After adult emergence in early February 2012, females were assigned to one of four feeding regimes—Good, Medium, Poor or Very Poor—for approximately six weeks. This increase in the time spent on feeding regimes (two weeks extra compared with [18]) was implemented to ensure that all females, particularly those in the Poor and Very Poor treatments, would have eggs in their ovaries at the time of mate choice experiments. Individuals in the Good treatment (N = 6) received four crickets three times per week, the Medium treatment (N = 6) received three crickets three times per week, the Poor treatment (N = 6) received two crickets three times per week and the Very Poor treatment (N = 6) received one cricket three times per week. Adult males continued on the juvenile feeding regime of two crickets three times per week. There was no significant difference in the body size (pronotum length) of females between treatments (ANOVA: F_3,23 = 0.604, p = 0.620; table 1), however the feeding treatments did create a significant difference in female body condition (ANOVA: F_3,23 = 24.336, p < 0.001; table 1) and female fecundity (ANOVA: F_3,23 = 37.578, p < 0.001; table 1). That is, female body condition and fecundity tended to increase as feeding treatment increased (see Tukey's in table 1). There was also a significant correlation between female body condition and fecundity across all groups (Pearson's: R = 0.838, N = 24, p < 0.001).

The fixed size and body condition of all females raised in the laboratory were similar to the range of female sizes and conditions previously recorded in nature (range [fixed size] = 13.19–19.13 mm, N = 22; range [body condition] = 0.016–0.072, N = 22), and the fixed size and body condition of males raised in the laboratory were similar to the range of male sizes and conditions previously recorded in nature (range [fixed size] = 11.830–15.160 mm, N = 27; range [body condition] = 0.015–0.022, N = 27) [29].

(d). Mate choice experiments

I carried out choice experiments in three large field enclosures (6 × 4 m and 3 m high) on the Macquarie University campus, North Ryde, Sydney, in mid-March 2012. The field enclosures were made of thick green plastic screen that allowed unobstructed airflow through each of the walls, but not through the roof. Ten small cylindrical cages (30 × 20 cm diameter)—two cages with females from each of the four feeding treatments and two empty control cages—were placed in a random order around the interior perimeter of each enclosure, and each enclosure was used only once (total females = 24, total controls = 6). So as to mimic the approximate natural sex ratio of 1 : 1 [28], eight virgin adult males were then arbitrarily chosen from the laboratory population and released onto foliage in the centre of each field enclosure (total starting males = 24). All of the small cages were covered in two layers of green plastic garden mesh to obscure visual cues while still allowing any chemical signals produced by the females to escape. The cages were checked for males at 07.00 each day over a 4-day period (males are most active at night/early morning; [18]) and any male found on a cage was removed from the field enclosure and replaced with a new virgin male from the laboratory population (also added to the central foliage). This procedure ensured that there were always eight males in each field enclosure. Fifty-five out of 78 males released into enclosures were found on female cages; the remaining 23 were either collected from the roof/walls of field enclosures at the end of the 4-day experimentation period or presumed dead and were therefore not included in analyses. The cumulative total of males found on a female's cage was used as an indicator of female attractiveness [18,20–22,34]. After the completion of these experiments, females were dissected and eggs counted under a stereomicroscope.

(e). Use of animals and fieldwork

No permits were obtained for the described field collections because New South Wales state law does not require specific permissions for the collection of invertebrates from locations outside of a national park. This study did not involve endangered or protected species.

(f). Data analysis

Data were analysed using SPSS 20.0 for Mac (SPSS Inc., Chicago, IL, USA) and were checked for normal distribution (Kolmogorov–Smirnov test) before further statistical analysis. Unless otherwise stated, all values are mean ± s.e. and all statistical tests are two-tailed. An ANOVA with Tukey's test was used to compare female characteristics (i.e. body size, body condition and fecundity) between the four treatment groups, and a Pearson's test was used for correlative analyses. A generalized linear model with Poisson distribution was used to compare the total number of males chemically attracted to each treatment group, with enclosure and day of experiment used as additional fixed factors. G-tests were subsequently used for pairwise comparisons and to compare overall male activity between enclosures and days. The size of fixed effects was estimated as the absolute difference in the total number of males as a proportion of the average number of males (for each pairwise comparison).

3. Results

Over the course of the experimental period, 18/24 females (75%) attracted a male. That is, 1/6 Good females (16.7%), 1/6 Medium females (16.7%), 3/6 Poor females (50%) and 1/6 Very Poor females (16.7%) were completely unattractive (G₃ = 1.726, p = 0.631). Every treatment group within an enclosure had at least one attractive female, and all females (N = 24) had at least one mature egg in their ovaries at the time of the mate choice experiments. No male was ever found on a control cage, so this treatment was removed from analyses. Overall male activity ranged from two to eight males per enclosure per day (mean number of males = 4.583 ± 0.543; electronic supplementary material, table S1), but did not significantly vary by enclosure (G₂ = 1.829, p = 0.401) or by day of experiment (G₃ = 1.462, p = 0.691).

Feeding treatment significantly affected female attractiveness (F_3,21 = 15.588, p = 0.001; figures 1 and 2), and there was no significant effect of enclosure (F_2,21 = 1.758, p = 0.415) or day of experiment (F_3,21 = 1.345, p = 0.718) on the number of males attracted to females. Further pairwise analyses showed at least a marginally significant difference in attractiveness between all treatment combinations except for between Good and Medium females (table 2). Across all feeding regimes, the total number of males attracted to each treatment group did not increase with increasing food quantity (figures 1 and 2) and the relationship between female body condition and attractiveness (R = −0.384, N = 24, p = 0.064) and female fecundity and attractiveness was not significant (R = −0.193, N = 24, p = 0.367). However, if Very Poor females were removed from these analyses, female attractiveness did indeed increase with increasing food quantity (figures 1 and 2) and a positive correlation between female body condition and attractiveness (R = 0.676, N = 18, p = 0.002) and fecundity and attractiveness emerged (R = 0.618, N = 18, p = 0.006).

Figure 1.

There was a significant difference in the attractiveness of females between treatment groups (F_3,21 = 15.588, p = 0.001). With the exception of Very Poor females, attractiveness (filled bars) increased with increasing body condition (unfilled bars). Most interestingly, females in the Very Poor treatment attracted significantly more males than females from any of the other treatment groups, even though they had the poorest body condition.

Figure 2.

There was a significant difference in the attractiveness of females between treatment groups (F_3,21 = 15.588, p = 0.001). With the exception of Very Poor females, attractiveness (filled bars) increased with increasing fecundity (unfilled bars). Most interestingly, females in the Very Poor treatment attracted significantly more males than females from any of the other treatment groups, even though they had the poorest fecundity.

Table 2.

Pairwise analysis of frequency data (number of males attracted to female treatments).

treatment combination	d.f.	G-value	p-value	effect size
Poor versus Very Poor	1	18.028	0.000022	1.47
Good versus Poor	1	5.884	0.015	1.11
Medium versus Poor	1	6.26	0.012	0.93
Medium versus Very Poor	1	6.26	0.012	0.81
Good versus Very Poor	1	3.656	0.056	0.6
Good versus Medium	1	0.361	0.548	0.24

4. Discussion

Feeding treatment created a sliding scale of body condition and fecundity for P. albofimbriata females: that is, an increase in food quantity translated to an increase in both body condition and fecundity. With the exception of the poorest-quality females, the number of males attracted to females increased with increasing female food quantity and there was a positive correlation between female body condition, fecundity and attractiveness, suggesting a system of overall signal honesty [2]. Most interestingly, however, females in the Very Poor feeding treatment attracted significantly more males in total than females from any of the other treatment groups even though Very Poor females were in the poorest body condition (figure 1) and had the lowest level of fecundity by far (figure 2). Furthermore, correlations between female body condition, fecundity and attractiveness were lost if Very Poor females remained in the analyses, suggesting signal dishonesty by the poorest-quality females. To my knowledge, this is the first study to provide evidence of intraspecific deception within the context of mate choice via chemical cues.

Although the cost of producing and releasing pheromone was not directly measured here, there is an increasing body of evidence suggesting that pheromones are indeed costly, affecting both survival and future reproduction (see [38] for review; [39,40], but see [41]). If signals were not costly for P. albofimbriata, we would expect all females to signal maximally, thereby increasing the chance of a meal and/or a mate. I suggest Very Poor females from this study are akin to Johnstone and Grafen's [2] ‘low-quality signallers with much to gain’, and females from the other treatments are either ‘high-quality signallers with less to gain’ (Good and Medium females) or ‘low-quality signallers with less to gain’ (Poor females). Very Poor females gain important resources from the easy meal attained as a result of deception—the consumption of one male improves body condition by approximately 33% and fecundity by approximately 40% [28]—so it is not surprising that females in the poorest condition invest maximally in pheromone production in order to increase the number of potential meals. Previous studies from non-cannibalistic taxa have also provided some support for the notion that poor condition induces an increase in signalling effort, most likely due to poor-condition individuals having a lower probability of survival and future reproduction than individuals in good condition, and therefore having more to gain from signalling at a higher level [42,43]. It is likely that P. albofimbriata females resort to cheating in only exceptional cases; that is, when their resource level is too low to survive a breeding cycle. Further study is required to determine where the specific switch point inducing deception lies in this species.

But the question still remains: how do the poorest-quality females manage to make a sufficient quality/quantity of pheromone if they have insufficient resources to make many eggs? As chemical cues have long been advocated as a relatively cheap signalling option [44], it may be that the resources required for pheromone production in P. albofimbriata are inexpensive compared with those required for egg production. In this case, the best strategy for the poorest-condition females would be to signal maximally and put off egg production until after a meal.

Comprehensive data on the proportion of these food-limited/poor-condition females in nature are yet to be collected for P. albofimbriata, but for the closely related and ecologically similar praying mantid Stagmomantis limbata, poor-quality females constitute only 3–9% of the mature population (depending on the time of season) [21]. Such a low frequency of poor-quality females suggests a generally reliable and honest system on average, and therefore any deceptive signalling within such a system is likely to be an ESS [2–5]. Another possibility is that males are able to uncover cheaters by basing their mate choice on multiple signals/traits [45,46]. Indeed, male mantids can visually perceive variation in female condition and fecundity and choose well-fed females in a simultaneous choice scenario [21,33]. However, they do not reject poorly fed females if they are the only option [28,47].

Unequivocal evidence of sexual deception implies a significant disadvantage to the receiver as a result of responding to the signal. Previous studies of P. albofimbriata found that females on the same diet as Very Poor females in this study are particularly high-risk for males—they are 90% likely to sexually cannibalize (as opposed to 0% for Good females; [28], and see table 1), only 50% of interactions result in copulation (as opposed to 100% in Good females; [28], and see table 1), and they have very few eggs to fertilize [18]. Taken together, these results suggest that females in the Very Poor treatment are deceptively signalling that they are in good condition and have an abundance of eggs when the reality for males is a twofold cost that includes a very high risk of cannibalism and low (or no) paternity. And as the overall signalling system is honest, males presumably fly to Very Poor females without any hesitation and have not evolved any obvious counter-adaptations to avoid these risky females [47].

Here, I provide the first empirical evidence in support of the Femme Fatale hypothesis—previous praying mantid mate choice experiments have revealed honest signalling systems in which poor-condition females are completely unattractive to males [18,20,21,34]. A system in which barren females are unable to properly signal their receptivity to males makes intuitive sense; however, results from this study tell a more complex story and predict that other systems might also discover cases of sexual deception if considering a broader range of female condition/fecundity. Although the underlying mechanism linking egg production and pheromone production in praying mantids is not well understood, the most plausible explanation is that both processes are under the same hormonal control, so that one cannot occur without the other [35,36,48]. Once egg production has begun, it is difficult to determine the underlying cause of differential attractiveness in praying mantids because feeding history, body condition and fecundity are generally correlated. Further studies that systematically attempt to tease apart these factors are needed if we are to determine the underlying mechanism responsible for pheromone production and subsequent female attractiveness.

Data accessibility

Mate choice data: Dryad Digital Repository doi:10.5061/dryad.h8m2q [49].

Funding statement

This project was funded by the Hermon Slade Foundation, the Association for the Study of Animal Behaviour and the Linnean Society of NSW.